surface climate over the North Atlantic during this century is associated with the global surface-warming trend during the 1920s and 1930s. The patterns of SST and air temperature change from 1900 to 1929 and from 1939 to 1968 (or, equivalently, the trends between 1917 and 1939) indicate that the warming was concentrated along the Gulf Stream east of Cape Hatteras. Warming also occurred over the Greenland Sea and the eastern subtropical Atlantic. The warming trend was accompanied by a decrease in the strength of the basin-scale atmospheric circulation (negative phase of the North Atlantic Oscillation). In marked contrast to the dipole EOF pattern, the wind changes occurred downstream of the largest SST anomalies. Hence the gradual surface warming along the Gulf Stream may have been a result of altered ocean currents rather than of local wind forcing.
A similar idea has been put forth to explain the cooling in the North Atlantic from the 1950s to the early 1970s (Levitus, 1989; Greatbach et al., 1991; Kushnir, 1994). The cooling trend was accompanied by a decrease in the transport of the Gulf Stream, as diagnosed from subsurface data by Greatbach et al. (1991). The decrease in Gulf Stream transport was in turn traced to bottom-pressure torque effects rather than wind effects (Greatbach et al., 1991).6
In a recent coupled atmosphere-ocean GCM experiment, Manabe and Stouffer (1988) showed that an intensified oceanic thermohaline circulation is associated with surface warming at high latitudes of the North Atlantic, and with an intensification and poleward shift of the Gulf Stream. The model's atmospheric response to an intensified thermohaline circulation is a weakening of the basin-wide wind system; this is consistent with the reduced baroclinicity. It is tempting to speculate that the climatic trends observed over the North Atlantic during the 1920s and 1930s were due to an intensification of the thermohaline circulation. However, we note that SSTs in the South Atlantic also warmed during this period (cf. Bottomley et al., 1990), contrary to the cooling observed in Manabe and Stouffer's experiment. Further modeling studies are needed to elucidate the role of the ocean circulation in low-frequency climate variability.
We thank Ludmilla Matrosova for her help with the data processing. We are indebted to Harry van Loon for stimulating discussions and to Michael Alexander for his thoughtful review of the manuscript. James Hurrell kindly provided the Bottomley et al. (1990) data set. This work was supported by a grant from NOAA's Climate and Global Change Program.
It should be noted that the spatial pattern of cooling during the 1950s to 1970s was somewhat different from the pattern of warming during the 1920s and 1930s (see Kushnir, 1994).
LAWRENCE A. MYSAK
Dr. Deser presented a very nice study. I enjoyed it, but I do not have a lot of comments at this stage. Some postwar data sets that have been analyzed show two to three cycles of interdecadal variability in the empirical orthogonal functions, for example. In Dr. Deser's analyses we have now seen decadal-scale variability over a 90-year record, which I for one really appreciate.
I think it is important to emphasize that only the winter data was analyzed. Dr. Deser and I discussed earlier whether this decadal variability shows up in other seasons, such as summer, and thus is a year-round signal. I think the answer is that it is there in the summer, but it is not as strong. It is definitely much more intense in winter. I should perhaps mention here that winter is defined as being from November to March, a five-month period.
The two time scales that I found particularly fascinating in the paper were the decadal and the biennial. While I will not discuss the biennial any further here, I just want to point out some recent work of one of our Ph.D. students at McGill, J. Wang, who also noticed a quasi-biennial cycle in the ice cover in the Labrador Sea and the Baffin Bay region. In addition, he found quite a strong correlation between anomalies in sea-ice extent and the North Atlantic oscillation (which is defined as a winter index).
The decadal time scale Dr. Deser discusses clearly falls between two familiar time scales: the North Atlantic oscillation, which has a period of 7 to 8 years, and the interdecadal cycles in the Arctic, which I shall discuss in my own presentation. Also, many recent modeling studies of the thermohaline circulation in the North Atlantic have shown the existence of internal oscillations with decadal to interdecadal periods.
I should like to list some relevant references, which I believe were not mentioned in the paper. Also, I might remark that many of these papers are by various people at this workshop. One piece of work is the master's thesis